Basic Reinforcement Learning Workflow¶

Define your target machine to run on¶

Import ROSE main modules:

from concurrent.futures import ProcessPoolExecutor

from radical.asyncflow import WorkflowEngine
from rhapsody.backends import ConcurrentExecutionBackend

from rose.metrics import GREATER_THAN_THRESHOLD
from rose.rl import SequentialReinforcementLearner

You can use and setup an HPC engine as we described in our previous Target Resources step:

engine = await ConcurrentExecutionBackend(ProcessPoolExecutor())
asyncflow = await WorkflowEngine.create(engine)

rl = SequentialReinforcementLearner(asyncflow)

Now our resource engine is defined, lets define our main RL workflow components:

@rl.environment_task
async def environment(*args):
    return f'python3 environment.py'

@rl.update_task
async def update(*args):
    return f'python3 update.py'

@rl.as_stop_criterion(metric_name='MODEL_REWARD', threshold=200, operator=GREATER_THAN_THRESHOLD)
async def check_reward(*args):
    return 'python3 check_reward.py'

Warning

For any metric function like @rl.as_stop_criterion the invoked script like check_reward.py must return a numerical value.

Finally invoke the tasks and register them with the reinforcement learner as a workflow:

# Start the RL training loop and break when stop condition is met
await rl.learn()

# You can also specify maximum iterations
await rl.learn(max_iter=10)
await asyncflow.shutdown()